A method for reducing power consumption in an active suspension system through the selective use of high performance, associated with high power demand, only in situations instantaneously deemed to provide a high ratio of benefit to cost. Input events are classified ahead of time, and are identified during operation of the system, ahead of time if possible through the use of look-ahead sensing or statistical analysis, or at the beginning of the event through the use of motion sensing. Once an event is detected, an estimation of the cost and benefits for an intervention of the active suspension system is made, and the intervention is scaled in a way to provide a good compromise. Relying on the nonlinearity of the cost and benefit expressions, this leads to overall reduced power consumption with small loss in perceived benefit.